Skip to main content
POST
Seal a value to propose later
Nothing is stored in the workspace’s secrets and nothing is proposed: an agent proposes the sealed value by giving the handle to its secrets_propose tool, so the value never passes through the model. The Reason CLI sends a command’s output here (<command> | reason secrets seal NAME). A handle works once, only for the name it was sealed for and only in this workspace, and expires after 60 minutes; the value is deleted when it is used or expires. What else a handle is bound to depends on the credential that sealed it. Sealed with an API key or a login (this operation’s documented credentials): the handle is a bearer handle within the workspace, bound to the workspace, the name and the member who sealed it, so only a Session turn acting for that member can propose it (seal_other_member otherwise), from any of their Sessions. Sealed inside a Session with the seal ticket secrets_propose gives the agent: the ticket seals once (a second use answers 409 seal_ticket_used), and the handle is also bound to that Session’s tree, so only that Session, its subagents and the Sessions they created can propose it (seal_other_session otherwise). The value is checked as a proposed secret is: the name must be shaped like a credential (invalid_secret_name), the value must not be empty, hold a NUL byte, exceed the size limit or be an address such as a URL or a connection string (secret_value_is_address), and a name already stored in the workspace answers 409 secret_exists. A sealed value is not masked in transcripts, logs or published text until it is stored as a secret: masking it earlier would let an agent test guesses about it. Seal a command’s output directly so the value is never printed. Set seen_by_model when the value already passed through an agent’s context: the approval card then recommends rotating it, and with session_id (a Session of the caller’s own) the exact value is replaced by a mask in that Session’s stored transcript; scrub reports what was rewritten. Requires secrets:write. The answer never contains the value. Scope: secrets:write

Authorizations

Authorization
string
header
required

Your Reason API key from Settings › API keys. New keys use reason_; legacy ara_ keys remain accepted. Keys are capability-scoped: run, mcp:read, mcp:write, secrets:read, secrets:write, sessions:read, sessions:debug, devices:read, devices:write, devices:use, computers:read, computers:write, ssh:write, knowledge:read, memory:read, memory:write, skills:read, skills:write, pages:read, pages:write, repos:read, repos:write, reviews:read, reviews:write, deployment:read, analytics:read, org:read, org:write, attachments:read, attachments:write, guardrails:read, guardrails:write, network:read, network:write, automations:read, automations:write, agent_auth:read. devices:write, devices:use, computers:read, computers:write, ssh:write, network:read and network:write are opt-in, excluded from default keys. computers:read lists visible cloud and connected computers without waking them or granting execution access. devices:read discovers local Device targets; devices:use separately consents to local execution, so a discover-and-run local workflow needs both. devices:write manages enrollment and device grants; computers:write creates or deletes cloud computers; ssh:write manages retained SSH access. network:read reads the saved workspace network policy; network:write replaces custom allowed hosts and requires a current organization owner/admin. Legacy project PATCH requests containing network_allowed_hosts also require network:write plus their existing run scope and project management permission; saved project hosts do not affect current session egress. Only owners/admins can mint or manage keys with network:write; keys cannot grant themselves more authority. See /docs/network-policies for replacement semantics and runtime applicability. mcp:write manages MCP server configuration only; it does not authorize remote MCP-tool execution. sessions:read reads sessions, including the assistant, reasoning and tool activity in their events. sessions:debug is privileged: only for organization owners/admins, it expands session events to the full diagnostic projection (diagnostic event kinds, status text and raw event metadata).

Scopes

secrets:write

Path Parameters

orgId
string
required

Organization id or slug. Resolve it with GET /v3/self.

Body

application/json
name
string
required

The name the value will be proposed under, shaped like a credential (NPM_TOKEN, STRIPE_API_KEY). The handle works for this exact name only.

value
string
required

The value to seal: not empty, not only whitespace, no NUL byte, not a URL or a connection string. Stored encrypted for at most an hour and never returned.

seen_by_model
boolean

Set when the value already passed through an agent's context (it was read or printed in a Session). The approval card then recommends rotating it.

session_id
string

With seen_by_model: the caller's own Session whose stored transcript holds the value, so the exact value is replaced by a mask there. It is where to look, nothing more; a Session that is not the caller's is left untouched.

Response

The value is sealed. handle is shown this once.